What is it?

The Work Assigner agent (Intelligent Workflow Assignment Agent) analyzes workflow tasks assigned to data steward users and redistributes them to keep the workload balanced. It identifies candidates by role or group, checks who is active based on recent activity, counts workflow tasks per user, and creates a redistribution plan that moves work away from inactive or overloaded users to active, underloaded ones.

Depending on your prompt, the agent can run in dry run mode (plan only) or execute changes, then generate a detailed Markdown-style report that shows the before-and-after task distribution and the actions it took.

For more information, see Work Assigner.

Redistribute workflow tasks by role or group

✅ Prompt: Redistribute workflow tasks for users with role ROLE_REVIEWER in tenant uK3BXQ7EecQZGBE. Balance the workload among active users.

Why it works: This prompt clearly specifies the role, the tenant, and the goal (“balance the workload among active users”). The agent can filter users by ROLE_REVIEWER, apply its default activity window (for example, last 7 days), and generate a redistribution plan that moves tasks from inactive or overloaded reviewers to active ones.

✅ Prompt: Rebalance tasks for ROLE_LOCAL_DS_US_DEV2. I want the difference between the busiest and least busy user to be no more than 2 tasks. Execute the changes.

Why it works: You provide the role, set a clear success criterion (maximum difference of two tasks), and explicitly request execution. The agent can set target_max_diff to 2, compute the before-and-after distribution, and then move tasks until the workload variance is within that tolerance.

✅ Prompt: Rebalance workflow tasks for GROUP_DATA_STEWARDS. Move tasks from users inactive for 7 or more days to active users and execute.

Why it works: This prompt uses a group identifier and defines the inactivity threshold explicitly. The agent can identify group members, mark those inactive based on the 7-day window, and reassign their tasks to active stewards.

Run dry runs before executing changes

✅ Prompt: I want to see a redistribution plan for GROUP_LOCAL_DS_US_DEV2. Don’t execute it yet — just show me what would happen. Consider users active if they’ve logged in within the last 14 days.

Why it works: You specify the group, clearly request a dry run, and customize the days_back parameter to 14 days instead of the default 7. The agent can generate a plan-only report with a before-and-after table and detailed actions without moving any tasks.

✅ Prompt: Find all users in ROLE_REVIEWER who are also in GROUP_DATA_STEWARDS. Redistribute their workflow tasks, moving all tasks from inactive users (no activity in 7 days) to active ones. Show me the plan first.

Why it works: This combines role and group filters and asks for a dry run. The agent builds an intersection of users, applies the 7-day activity threshold, and shows you how tasks would shift before you choose to execute.

Handle emergencies and edge cases

✅ Prompt: User john.smith is on leave. Reassign all his workflow tasks to other active users in ROLE_REVIEWER. Execute immediately.

Why it works: You provide the user who should lose tasks, define the candidate pool (other active reviewers), and request execution. The agent can treat this as an emergency rebalance and move all tasks off john.smith while keeping the new distribution as balanced as possible.

✅ Prompt: Show me current task distribution for ROLE_REVIEWER. Don’t make any changes — just report who has how many tasks and who is active.

Why it works: This prompt asks for an audit-only view. The agent can list each reviewer with task counts, percentages, and active or inactive status based on recent activity without proposing or executing any redistribution.

✅ Prompt: Dry run only: redistribute tasks for ROLE_REVIEWER. I want to see if new user jane.doe will get tasks assigned.

Why it works: You call out a specific scenario (onboarding a new user) and request a dry run. The agent can simulate the redistribution, show how many tasks would go to jane.doe, and help you decide whether to proceed.

Avoid ambiguous or out-of-scope prompts

⚠️ Prompt: Balance the workflow tasks.

Issue: This doesn’t specify which users to consider (no role or group) and doesn’t say whether you want a dry run or execution. The agent can’t safely choose on its own.

Better prompt: Redistribute workflow tasks for users with role ROLE_REVIEWER. Show me a plan first.

⚠️ Prompt: Redistribute tasks for ROLE_USER but only to people who are really active.

Issue: “Really active” is subjective. The agent has a default activity window (for example, 7 days), but your expectation might be 1 day or 30 days.

Better prompt: Redistribute tasks for ROLE_USER, considering users active if they logged in within the last 14 days.

⚠️ Prompt: Show me a dry run for ROLE_REVIEWER, but also execute the changes and send me a report.

Issue: This mixes two conflicting instructions. A dry run means “don’t execute,” but you also ask to execute. The agent must pick one behavior or ask for clarification.

Better prompt: Show me a dry run for ROLE_REVIEWER only. I’ll confirm if I want to execute afterward.

⚠️ Prompt: Redistribute tasks for ROLE_REVIEWER and also update their user profiles to set their department to “Data Quality.”

Issue: Profile updates are out of scope for this agent. It only handles workflow task redistribution; user profile changes require separate tools or processes.

Better prompt: Redistribute tasks for ROLE_REVIEWER only. I’ll handle profile updates separately.

⚠️ Prompt: Move tasks from the inactive people to the active ones.

Issue: The agent doesn’t know which “people” to consider and how inactivity should be defined. It needs at least a role or group and, ideally, an explicit activity window.

Better prompt: Move tasks from inactive users to active users in GROUP_DATA_STEWARDS, considering users inactive if they haven’t logged in for 30 days.

Best practices